Institute of Work Health and Safety / Reporting

Low incident numbers are ambiguous

They can mean a safe operation or a quiet one, and the two look identical on a chart.

A falling injury rate is treated as unambiguously good and is not. Fewer recorded incidents can mean the work became safer. It can also mean people stopped reporting, which happens reliably whenever reporting starts to carry a cost, and the cost need not be formal. An investigation that focuses on what the injured person did wrong teaches everybody watching what reporting produces.

The two explanations can be distinguished, but not from the injury numbers alone. Near miss reports moving in the opposite direction to injuries is a healthy signal, because it suggests people are surfacing things before they cause harm. Both falling together is worth being suspicious about, particularly if it coincides with a period of production pressure or a change in how incidents are handled.

This is why the metrics that count things that have already gone wrong are a poor steering instrument even though they are the ones on the board report. They are accurate about the past and silent about the present, and they can be improved by discouraging reports, which is the one improvement nobody wants.

The organisations that handle this well tend to watch the reporting rate itself as a measure of system health, and to treat a sudden drop as a question rather than an achievement.

There is a further complication that makes the numbers even harder to read, which is that the definition of a recordable event is itself a decision. Where the boundary sits between a first aid case and a recordable injury involves judgement, and that judgement is exercised by people who know how the figure will be used. No dishonesty is required for the boundary to drift over a period when the number matters, and the drift is invisible in the data because the data is the thing being defined.